Computational methods atomistic simulation

Very recently, people who engage in computer simulation of crystals that contain dislocations have begun attempts to bridge the continuum/atomistic divide, now that extremely powerful computers have become available. It is now possible to model a variety of aspects of dislocation mechanics in terms of the atomic structure of the lattice around dislocations, instead of simply treating them as lines with macroscopic properties (Schiotz et al. 1998, Gumbsch 1998). What this amounts to is linking computational methods across different length scales (Bulatov et al. 1996). We will return to this briefly in Chapter 12. 

From these early beginnings, computer studies have developed into sophisticated tools for the understanding of defects in solids. There are two principal methods used in routine investigations atomistic simulation and quantum mechanics. In simulation, the properties of a solid are calculated using theories such as classical electrostatics, which are applied to arrays of atoms. On the other hand, the calculation of the properties of a solid via quantum mechanics essentially involves solving the Schrodinger equation for the electrons in the material. 

There are two other methods in which computers can be used to give information about defects in solids, often setting out from atomistic simulations or quantum mechanical foundations. Statistical methods, which can be applied to the generation of random walks, of relevance to diffusion of defects in solids or over surfaces, are well suited to a small computer. Similarly, the generation of patterns, such as the aggregation of atoms by diffusion, or superlattice arrays of defects, or defects formed by radiation damage, can be depicted visually, which leads to a better understanding of atomic processes. 

Although a first principle or ab initio atomistic simulation of a one-million atom system is being attempted with one of the most powerful computers, ab initio atomistic modeling of a macroscopic (1024 atoms) in a long time scale (103 s) is seemly not possible in a foreseeable future. The hierarchical multiscale simulation methods are implementable options for the time being. 

The prerequisites which make possible the development of atomistic simulations of diffusion in polymers are the development of powerful methods for the simulation of polymer microstructures and dynamics and also great computation capabilities of supercomputers. 

On the other hand, based on the rapid progress which was recorded in the last decade in the atomistic simulation of diffusion processes in polymers one may be confident that these computational methods will be one day able to cope with the prob-... 

The periodic approach is not the only one available for atomistic simulations of these materials and we should first mention that much progress has been made in the application of molecular quantum chemical methods using cluster representations of the local structure of oxide materials [1, 2], More recently, this has given way to mixed quantum mechanics/molecular mechanics (QM/MM) calculations. In QM/MM simulations the important region, the active site for catalysis, is represented at a quantum chemical level while the influence of its environment, the extended solid, is represented using the computationally less-demanding atomistic force field approach. This allows complex structures such as metal particles supported on oxides to be tackled [3]. 

Jayaraman and Maginn calculated two crystal polymorphs of [C4mim][Cl] using a thermodynamic integration-based atomistic simulation method [78]. The computed... 

Empirical correlations [21] and computationally-intensive large-scale atomistic simulations [27-42] have been mentioned above as opposite extremes in the spectrum of methods used to treat diffusion phenomena. In addition, two general types of phenomenological theories (which are much more sophisticated than empirical correlations, but much more coarse-grained than atomistic simulation methods) have been developed to treat diffusion ... 

In addition to the classical force fields above, many other force fields have been developed for small drug molecules or macromolecules. The MM2, MM3, and MM4 force fields were developed by Norman L. Allinger for a broad range of chemicals, and CFF is a family of force fields adapted to a broad variety of organic compounds, polymers, metals, and so on. The MMFF force field was developed at Merck for a broad range of chemicals. ReaxFF is a reactive force field, developed by William Goddard and coworkers, is fast, transferable, and the computational method of choice for atomistic-scale dynamics simulations of chemical reactions. 

Markus Meuwly is professor of physical and computational chemistry at the Department of Chemistry of the University of Basel and adjunct research professor at Brown University, USA. He is interested in developing computational/theoretical methods for quantitative atomistic simulations, specifically multipolar force fields and reactive processes in complex systems. 

---